skip to navigation
skip to content

property-manager 2.2

Useful property variants for Python programming (required properties, writable properties, cached properties, etc)

The property-manager package defines several custom property variants for Python programming including required properties, writable properties, cached properties, etc. It’s currently tested on Python 2.6, 2.7, 3.4, 3.5, 3.6 and PyPy. For usage instructions please refer to the documentation.


The property-manager package came into existence as a submodule of my executor package where I wanted to define classes with a lot of properties that had a default value which was computed on demand but also needed to support assignment to easily override the default value.

Since I created that module I’d wanted to re-use it in a couple of other projects I was working on, but adding an executor dependency just for the property_manager submodule felt kind of ugly.

This is when I decided that it was time for the property-manager package to be created. When I extracted the submodule from executor I significantly changed its implementation (making the code more robust and flexible) and improved the tests, documentation and coverage in the process.


The property-manager package is available on PyPI which means installation should be as simple as:

$ pip install property-manager

There’s actually a multitude of ways to install Python packages (e.g. the per user site-packages directory, virtual environments or just installing system wide) and I have no intention of getting into that discussion here, so if this intimidates you then read up on your options before returning to these instructions ;-).


This section shows how to use the most useful property subclasses. Please refer to the documentation for more detailed information.

Writable properties

Writable properties with a computed default value are easy to create using the writable_property decorator:

from random import random
from property_manager import writable_property

class WritablePropertyDemo(object):

    def change_me(self):
        return random()

First let’s see how the computed default value behaves:

>>> instance = WritablePropertyDemo()
>>> print(instance.change_me)
>>> print(instance.change_me)

As you can see the value is recomputed each time. Now we’ll assign it a value:

>>> instance.change_me = 42
>>> print(instance.change_me)

>From this point onwards change_me will be the number 42 and it’s impossible to revert back to the computed value:

>>> delattr(instance, 'change_me')
Traceback (most recent call last):
  File "property_manager/", line 584, in __delete__
    raise AttributeError(msg % (obj.__class__.__name__, self.__name__))
AttributeError: 'WritablePropertyDemo' object attribute 'change_me' is read-only

If you’re looking for a property that supports both assignment and deletion (clearing the assigned value) you can use mutable_property.

Required properties

The required_property decorator can be used to create required properties:

from property_manager import PropertyManager, required_property

class RequiredPropertyDemo(PropertyManager):

    def important(self):
        """A very important attribute."""

What does it mean for a property to be required? Let’s create an instance of the class and find out:

>>> instance = RequiredPropertyDemo()
Traceback (most recent call last):
  File "property_manager/", line 131, in __init__
    raise TypeError("%s (%s)" % (msg, concatenate(missing_properties)))
TypeError: missing 1 required argument (important)

So the constructor of the class raises an exception when the property hasn’t been given a value. We can give the property a value by providing keyword arguments to the constructor:

>>> instance = RequiredPropertyDemo(important=42)
>>> print(instance)

We can also assign a new value to the property:

>>> instance.important = 13
>>> print(instance)

Cached properties

Two kinds of cached properties are supported, we’ll show both here:

from random import random
from property_manager import cached_property, lazy_property

class CachedPropertyDemo(object):

    def expensive(self):
        print("Calculating expensive property ..")
        return random()

    def non_idempotent(self):
        print("Calculating non-idempotent property ..")
        return random()

The properties created by the cached_property decorator compute the property’s value on demand and cache the result:

>>> instance = CachedPropertyDemo()
>>> print(instance.expensive)
Calculating expensive property ..
>>> print(instance.expensive)

The property’s cached value can be invalidated in order to recompute its value:

>>> del instance.expensive
>>> print(instance.expensive)
Calculating expensive property ..
>>> print(instance.expensive)

Now that you understand cached_property, explaining lazy_property is very simple: It simply doesn’t support invalidation of cached values! Here’s how that works in practice:

>>> instance.non_idempotent
Calculating non-idempotent property ..
>>> instance.non_idempotent
>>> del instance.non_idempotent
Traceback (most recent call last):
  File "property_manager/", line 499, in __delete__
    raise AttributeError(msg % (obj.__class__.__name__, self.__name__))
AttributeError: 'CachedPropertyDemo' object attribute 'non_idempotent' is read-only
>>> instance.non_idempotent

Properties based on environment variables

The constructor of the custom_property class (and its subclasses) accepts the keyword argument environment_variable which can be provided to get the property’s value from the environment:

from random import random
from property_manager import mutable_property

class EnvironmentPropertyDemo(object):

    def environment_based(self):
        return 'some-default-value'

By default the property’s value is computed as expected:

>>> instance = EnvironmentPropertyDemo()
>>> print(instance.environment_based)

When the environment variable is set it overrides the computed value:

>>> os.environ['WHATEVER_YOU_WANT'] = '42'
>>> print(instance.environment_based)

Assigning a value to the property overrides the value from the environment:

>>> instance.environment_based = '13'
>>> print(instance.environment_based)

Deleting the property clears the assigned value so that the property falls back to the environment:

>>> delattr(instance, 'environment_based')
>>> print(instance.environment_based)

If we now clear the environment variable as well then the property falls back to the computed value:

>>> os.environ.pop('WHATEVER_YOU_WANT')
>>> print(instance.environment_based)

Support for setters and deleters

All of the custom property classes support setters and deleters just like Python’s property decorator does.

The PropertyManager class

When you define a class that inherits from the PropertyManager class the following behavior is made available to your class:

  • Required properties raise an exception if they’re not set.
  • The values of writable properties can be set by passing keyword arguments to the constructor of your class.
  • The repr() of your objects will render the name of the class and the names and values of all properties. Individual properties can easily be excluded from the repr() output.
  • The clear_cached_properties() method can be used to invalidate the cached values of all cached properties at once.

Similar projects

The Python Package Index contains quite a few packages that provide custom properties with similar semantics:

My personal favorite until I wrote my own :-). This package provides several cached property variants. It supports threading and time based cache invalidation which property-manager doesn’t support.
This package provides two cached property variants: a read only property and a writable property. Both variants cache computed values indefinitely.
This package provides a single property variant which simply caches computed values indefinitely.
This package provides several cached property variants supporting class properties, object properties and cache invalidation.
This package uses metaclasses to implement an alternative syntax for defining computed properties. It defines several property variants with semantics that are similar to those defined by the property-manager package.
This package implements computed, writable properties using an alternative syntax to define the properties.

Distinguishing features

Despite all of the existing Python packages discussed above I decided to create and publish the property-manager package because it was fun to get to know Python’s descriptor protocol and I had several features in mind I couldn’t find anywhere else:

  • A superclass that sets writable properties based on constructor arguments.

  • A superclass that understands required properties and raises a clear exception if a required property is not properly initialized.

  • Clear disambiguation between lazy properties (whose computed value is cached but cannot be invalidated because it would compromise internal state) and cached properties (whose computed value is cached but can be invalidated to compute a fresh value).

  • An easy way to quickly invalidate all cached properties of an object.

  • An easy way to change the semantics of custom properties, e.g. what if the user wants a writable cached property? With property-manager it is trivial to define new property variants by combining existing semantics:

    from property_manager import cached_property
    class WritableCachedPropertyDemo(object):
        def expensive_overridable_attribute(self):
            """Expensive calculations go here."""

    The example above creates a new anonymous class and then immediately uses that to decorate the method. We could have given the class a name though:

    from property_manager import cached_property
    writable_cached_property = cached_property(writable=True)
    class WritableCachedPropertyDemo(object):
        def expensive_overridable_attribute(self):
            """Expensive calculations go here."""

    By giving the new property variant a name it can be reused. We can go one step further and properly document the new property variant:

    from property_manager import cached_property
    class writable_cached_property(cached_property):
        """A cached property that supports assignment."""
        writable = True
    class WritableCachedPropertyDemo(object):
        def expensive_overridable_attribute(self):
            """Expensive calculations go here."""

    I’ve used computed properties for years in Python and over those years I’ve learned that different Python projects have different requirements from custom property variants. Defining every possible permutation up front is madness, but I think that the flexibility with which the property-manager package enables adaptation gets a long way. This was the one thing that bothered me the most about all of the other Python packages that implement property variants: They are not easily adapted to unanticipated use cases.


The latest version of property-manager is available on PyPI and GitHub. The documentation is hosted on Read the Docs. For bug reports please create an issue on GitHub. If you have questions, suggestions, etc. feel free to send me an e-mail at


This software is licensed under the MIT license.

© 2017 Peter Odding.

File Type Py Version Uploaded on Size
property-manager-2.2.tar.gz (md5) Source 2017-06-29 24KB
property_manager-2.2-py2.py3-none-any.whl (md5) Python Wheel py2.py3 2017-06-29 24KB